Multimodal Large Language Models (MLLMs) have demonstrated impressive capabilities in both visual understanding and code generation. However, existing benchm…
机构:上交
来源:arXiv 2610.10066 | AI4Papers 论文推荐平台